Case Studies

Real-world results from running Nerviq across different stacks, platforms, and team configurations.

Single PlatformClaude Code

Typical Next.js SaaS

12 → 56score

nerviq setup delivered +44 points on a greenfield Next.js repo in one pass.

Single PlatformClaude Code

Stripe Integration App

12 → 57score

Payment-adjacent repos benefit sharply from explicit trust and secret-handling checks.

Single PlatformClaude Code

Mobile App

12 → 55score

Mobile stacks need explicit build/run documentation for agent reliability.

Single PlatformClaude Code

Turborepo Monorepo

12 → 52score

Monorepos score lower without root-level agent map and package boundaries.

Single PlatformClaude Code

Node CLI Package

14 → 57score

CLI projects gain from documented release and test entrypoints.

Single PlatformClaude Code

FastAPI Backend

50score

Python services start higher (~50) thanks to opinionated tooling and layouts.

Single PlatformClaude Code

Django Web App

49score

Django conventions help scores but agent instructions often omit deployment context.

Single PlatformClaude Code

Go Microservice

45score

Go repos benefit from explicit module boundaries in instructions.

Single PlatformClaude Code

Rust Library

47score

Rust projects score well on hygiene; tooling docs close the last mile.

Single PlatformClaude Code

LangChain / RAG Project

50score

ML repos need explicit data-handling and evaluation instructions for safe agents.

Single PlatformClaude Code

Claudex Research Engine (live)

85/100score

Highly configured research repo; intentional gaps in container/IaC signals.

Single PlatformClaude Code

Nerviq CLI — Claude audit (live)

83/100score

Self-audit: nerviq meets its own Claude Code bar with zero critical issues.

Single PlatformCodex

Nerviq CLI — Codex uplift (live)

35 → 84score

+49 points after setup proves Codex path parity with deliberate AGENTS.md work.

Single PlatformCodex

Cross-platform self-check (Codex)

82/100score

82 Codex vs 83 Claude — near-identical alignment when AGENTS.md anchors intent.

HarmonyClaude CodeCodex

Claude + Codex on one repo

Harmony 84score

One critical Codex CI auth gap; otherwise strong two-platform alignment.

HarmonyClaudeCursorCopilot

Claude + Cursor + Copilot

Harmony 84score

Harmony moved 42→84 after introducing AGENTS.md as shared anchor.

HarmonyAll 8

All eight platforms

Harmony 81score

Eight platforms is viable; security drift is the dominant risk vector.

EnterpriseCursorClaudeCopilot

10-person TypeScript monorepo

Harmony 83score

+42 harmony points in six weeks; ~65 engineer-hours invested.

EnterpriseClaudeAiderCopilot

200-contributor OSS library

Harmony 73score

Contributor heterogeneity makes AGENTS.md non-optional.

EnterpriseClaudeCursorAider

Two-founder MVP (8 weeks)

Harmony 74score

~3 hour setup investment → +40–45 points within two weeks.

EnterpriseClaudeCursorCopilot

PCI + SOC 2 payments platform

Harmony 87score

Compliance forcing functions produced the highest harmony in the corpus.

EnterpriseClaudeAiderCopilot

Terraform + Kubernetes platform

Harmony 76score

Pre-tool-use hooks and blast-radius docs matter more than generic linting.

MigrationCursorClaude

Cursor → Claude expansion

Harmony 31 → 85score

.cursorrules silently ignored in Agent mode until nerviq surfaced it.

SynergyMulti-platform

90-day controlled synergy study

+45% velocityscore

High harmony (80+) multiplies value; low harmony barely beats single-platform.

SynergyAll

24-study cross-cutting synthesis

Patternsscore

AGENTS.md appears in 18/24 studies as the primary harmony driver.